期刊文献+

基于决策树算法构建急性胰腺炎肠外营养患者静脉导管相关血流感染预测模型

Construction of a predictive model for venous catheter-related bloodstream infection in patients with acute pancreatitis undergoing parenteral nutrition based on decision tree algorithm
暂未订购
导出
摘要 目的探讨急性胰腺炎(AP)肠外营养患者静脉导管相关血流感染(CRBSI)危险因素,构建决策树预测模型并验证。方法回顾性分析2021年1月至2023年8月九江市某医院收治的AP肠外营养患者140例为建模组,按照患者是否发生CRBSI分为CRBSI组20例和非CRBSI组120例。采用单因素和二元logistic回归分析AP肠外营养患者发生CRBSI的影响因素,构建决策树预测模型。另按7∶3分配比选取2023年9月至2024年4月该院AP肠外营养患者60例为验证组。采用ROC曲线评估模型预测效能。结果建模组140例AP肠外营养患者CRBSI的发生率14.29%。多因素分析结果显示年龄≥60岁、肠外营养时间>10 d、置管持续时间>14 d、合并糖尿病均是AP肠外营养患者发生CRBSI的独立影响因素(P<0.05)。根据发生CRBSI所占节点构成比代表AP肠外营养患者发生CRBSI的风险,筛选出4类高危人群:①肠外营养时间>10 d、合并糖尿病,占该节点的62.50%;②肠外营养时间>10 d、不合并糖尿病、置管持续时间>14 d,占该节点的38.50%;③肠外营养时间≤10 d、年龄≥60岁、置管持续时间>14 d,占该节点的40.00%;④肠外营养时间≤10 d、年龄≥60岁、置管持续时间≤14 d、合并糖尿病,占该节点的20.00%。ROC曲线中,训练集的曲线下面积(AUC)为0.929(95%CI:0.877~0.980),灵敏度为80.00%,特异度为87.50%;验证集的AUC为0.926(95%CI:0.840~1.000),灵敏度为85.00%,特异度为83.00%。结论根据AP肠外营养患者发生CRBSI的相关影响因素构建的决策树预测模型,预测效果良好,临床医护工作者或可根据此模型筛选出易发生CRBSI的高风险AP肠外营养患者,从而及时开展干预措施。 Objective To explore the risk factors for intravenous catheter-related blood-stream infection(CRBSI)in pa-tients with acute pancreatitis(AP)with parenteral nutrition,and to construct a decision tree prediction model with validation.Methods A retrospective analysis was conducted on 140 AP patients receiving parenteral nutrition admitted to a Hospital of Jiu-jiang from January 2021 to August 2023,who were included in the modeling group.They were divided into the CRBSI group(20 cases)and the non-CRBSI group(120 cases)according to whether CRBSI occurred.Univariate analysis and binary logistic re-gression analysis were used to identify the influencing factors of CRBSI in AP patients receiving parenteral nutrition,and a deci-sion tree prediction model was constructed.In addition,60 AP patients receiving parenteral nutrition in the hospital from Septem-ber 2023 to April 2024 were selected as the validation group according to the 7∶3 allocation ratio.The receiver operating charac-teristic(ROC)curve was used to evaluate the predictive efficacy of the model.Results The incidence of CRBSI in 140 AP pa-tients receiving parenteral nutrition in the modeling group was 14.29%.Multivariate analysis showed that age≥60 years old,parenteral nutrition duration>10 days,catheter indwelling time>14 days,and complicated with diabetes were independent risk factors for CRBSI in AP patients receiving parenteral nutrition(P<0.05).According to the proportion of nodeswith CRBSI,which represents the risk of CRBSI in AP patients receiving parenteral nutrition,4 types of high-risk populations were screened out:①parenteral nutrition duration>10 days and complicated with diabetes,accounting for 62.50%of the node;②parenteral nutrition duration>10 days,not complicated with diabetes,and catheter indwelling time>14 days,accounting for 38.50%of the node;③parenteral nutrition duration≤10 days,age≥60 years old,and catheter indwelling time>14 days,accounting for 40.00%of the node;④parenteral nutrition duration≤10 days,age≥60 years old,catheter indwelling time≤14 days,and complicated with diabetes,accounting for 20.00%of the node.In the ROC curve,the area under the curve(AUC)of the train-ing set was 0.929(95%CI:0.877-0.980),with a sensitivity of 80.00%and a specificity of 87.50%;the AUC of the valida-tion set was 0.926(95%CI:0.840-1.000),with a sensitivity of 85.00%and a specificity of 83.00%.Conclusion The de-cision tree prediction model was constructed according to the relevant influencing factors of the occurrence of CRBSI in patients with AP parenteral nutrition,and the prediction effect was good.Clinical health care workers may screen out the high-risk pa-tients with AP parenteral nutrition who are prone to the occurrence of CRBSI according to the model,so as to carry out timely in-tervention measures.
作者 柯四水 廖静 Ke Sishui;Liao Jing(Jiujiang First People′s Hospital,Jiujiang 332000,China)
出处 《中国医院统计》 2026年第1期36-42,共7页 Chinese Journal of Hospital Statistics
关键词 急性胰腺炎 静脉导管 血流感染 决策树 预测模型 acute pancreatitis intravenous catheter bloodstream infection decision tree prediction model
  • 相关文献

参考文献18

二级参考文献155

共引文献350

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部